/
Navigation
Chronicles
Browse all articles
Explore
Semantic exploration
Research
Entity momentum
Nexus
Correlations & relationships
Story Arc
Topic evolution
Drift Map
Semantic trajectory animation
Posts
Analysis & commentary
Pulse API
Tech news intelligence API
Browse
Entities
Companies, people, products, technologies
Domains
Browse by publication source
Handles
Browse by social media handle
Detection
Concept Search
Semantic similarity search
High Impact Stories
Top coverage by position
Sentiment Analysis
Positive/negative coverage
Anomaly Detection
Unusual coverage patterns
Analysis
Rivalry Report
Compare two entities head-to-head
Semantic Pivots
Narrative discontinuities
Crisis Response
Event recovery patterns
Connected
Search: /
Command: ⌘K
Embeddings: large
TEXXR

Chronicles

The story behind the story

days · browse · Enter similar · o open

An overview of macro tech trends for 2025, focusing on generative AI, LLMs, scaling challenges with training ever bigger AI models, the capex surge, and more

Presentations Every year, I produce a big presentation exploring macro and strategic trends in the tech industry.  For 2025, ‘AI eats the world’.

Benedict Evans

Context & Ripple Effects

This annual macro-tech series had previously moved from the still-early cloud transition and Web3/metaverse themes to generative-AI opportunities beyond text and images. The 2025 edition concentrates that arc on LLMs and the infrastructure required to advance them.

Its “AI eats the world” framing makes model scaling and capital spending a strategic question rather than a single-product trend. Later coverage’s emphasis on chip demand, supply constraints and model commoditization shows why this framing matters: AI investment can reshape both product competition and the supply base.

First-order effects

  • The presentation directs attention toward generative AI and LLMs as near-term product and research priorities, while making the cost and difficulty of training larger models a central constraint.
  • Expected AI capex shifts immediate strategic focus to compute and other infrastructure capacity, not just end-user AI features.

Second-order effects

  • Companies competing on AI products face pressure to pair model ambitions with a credible infrastructure strategy; scaling constraints can favor players able to fund or secure capacity.
  • A capex-led race increases the importance of compute economics and can shift differentiation toward deployment, distribution and product integration as models become less exclusive.

Third-order effects

  • If persistent, the pattern points to AI industrialization: competitive advantage will be determined across a stack of models, compute and applications, rather than by model quality alone.
  • The scale of required investment could further concentrate frontier-model development among firms with capital and infrastructure access, though the eventual effect depends on whether scaling continues to produce sufficiently valuable gains.

The trend: AI is moving from a software-feature cycle toward an infrastructure-intensive platform shift in which model progress, compute availability and commercialization are tightly linked.

Discussion

  • @benedictevans Benedict Evans on threads
    Every year, I produce a big presentation exploring macro and strategic trends in the tech industry.  For 2025, ‘AI eat the world’.  More questions than answers, but such interesting questions. https://www.ben-evans.com/...
  • @bogmoroz Bogdan on threads
    I was lucky to catch this in person at @slushhq yesterday—packed with great insights, as always!
  • @petemarcus Pete Marcus on threads
    From a comms perspective, I wonder if this slide from @benedictevans' presentation is the most important?  The business model for the past 20 years of the consumer internet is changing, and the comms approach will need to as well...
  • @glinden.bsky.social Greg Linden on bluesky
    I think the biggest question is whether smaller, cheaper LLMs are good enough for almost all valuable use cases.  Unless many need the expensive huge models to do something a lot of people find very valuable, a lot of business models are going to collapse.  This talk dances aroun…
  • @cuthrell.com Jay Cuthrell on bluesky
    The analogous reference infinite interns i.e. Abundant Interns (AI) works on several levels.  “Free” internships give way to paid internships — and companies that succeed with interns typically have invested in structure, planning, mentorship, and integration.
  • @coolshannes Hannes Cools on x
    Highly recommended 🧐 Every year, @benedictevans produces a presentation exploring macro and strategic trends in the tech industry. For 2025, his presentation is titled ‘AI eats the world’. More questions than answers, but such interesting questions! 👇 https://www.ben-evans.com/..…
  • @_ram_ Ram Parameswaran on x
    A classic slide from @benedictevans https://www.ben-evans.com/... [image]
  • @stevesi Steven Sinofsky on x
    Benedict Evans puts together a yearly slide deck looking broadly at the industry while diving in with specific numbers and a lot of charts (90 slides). It is always a gem. via @benedictevans This year is about AI (of course). It is super. https://www.ben-evans.com/... [image]
  • @nikmilanovic Nik on x
    The whole @benedictevans gen AI writeup is interesting, but I liked this chart in particular. https://www.ben-evans.com/... We're very much still in the bolt-on era. Reminds me of this article: “Why didn't electricity immediately change manufacturing?” https://www.bbc.com/... [im…